workaround for vector based significance level

examples/mixture_normal_var.jl

@@ -3,11 +3,14 @@ using MarketRisk
 ω = [0.75, 0.25]
 μ = [0.0, -0.1]
 Σ = [0.2, 0.4]
-αs = 0.001
+α = 0.001
+αs = [0.001, 0.01, 0.1]
 h = 10 / 250
 
 m = MixtureModel(Normal, [(h * μ[i], sqrt(h) * Σ[i]) for i in 1:2], ω)
 
-mixture_var = ValueAtRisk(μ, Σ, h, αs, m)
-compute(mixture_var)
+mixture_var = ValueAtRisk(μ, Σ, h, α, m)
+mixture_var1 = ValueAtRisk(μ, Σ, h, αs, m)
 
+compute(mixture_var)
+compute(mixture_var1)

src/value_at_risk.jl

@@ -26,7 +26,21 @@ end
 function compute(var::ParametricValueAtRisk{V, M, H, A, MixtureModel{B, D, Normal, F}}) where {V, M, H, A, B, D, F}
  @unpack μ, Σ, h, α, dist = var 
 
+ opt(x, α) = cdf(dist, x) - α
  D_opt(f) = x -> ForwardDiff.derivative(f, float(x))
 
- @. - find_zero((x -> cdf(dist, x) - α, D_opt(x -> opt_func(x, α))), 0.0, Roots.Newton())
+ - find_zero((x -> opt(x, α), D_opt(x -> opt(x, α))), 0.0, Roots.Newton())[1]
 end
+
+function compute(var::ParametricValueAtRisk{V, M, H, A, MixtureModel{B, D, Normal, F}}) where {V, M, H, A<:Vector{<:Real}, B, D, F}
+ @unpack μ, Σ, h, α, dist = var 
+
+ opt(x, α) = cdf(dist, x) - α
+ D_opt(f) = x -> ForwardDiff.derivative(f, float(x))
+
+ v = zeros(eltype(α), length(α))
+ for i in 1:length(α) 
+ v[i] = - find_zero((x -> opt(x, α[i]), D_opt(x -> opt(x, α[i]))), 0.0, Roots.Newton())[1]
+ end
+ v
+end